Skip to main content
T0 Big tech Alibaba Round 1

Alibaba AI Infra first round (2)

Real Alibaba AI Infra first-round interview questions covering high-performance computing and related topics

High-performance computing

Fundamentals

  1. Describe the full Qwen-VL training pipeline.
  2. In multimodal LLMs, what common ways pass visual features into the LLM?
  3. How is ViT usually pretrained?
  4. For images or video of different sizes, how should positional encoding be designed?
  5. What do residual connections do in ResNet or Transformer? Expand on it.
  6. Introduce the basic idea and implementation of multimodal RAG.

Project experience

  1. Introduce yourself and walk through project experience in detail.
  2. Outline the large-model training pipeline.

Open questions

  1. When GPU memory is short during training or inference, what optimization options exist?

Coding

  1. Kth largest element in an array (LeetCode 215).